Monitoring particulate pollution using GOCI COMS

Transcription

1 CEOS AC-VC Meeting June, 2017, CNES, Paris Monitoring particulate pollution using GOCI COMS COMS/GOCI&MI Jhoon Kim 1, M. Choi 1, H. Lim 1, Tom Eck, Brent Holben, Zhengqiang Li 1 Yonsei University, 2 NASA GSFC, 3 IRSA, CAS 1

2 Collaborators National Institute of Environmental Research(NIER) Ji-Hyung Hong, You Deog Hong, Lim-Seok Chang, Joon-Young Ahn, Jeong-Hoo Park Sites in Korea Baeksa Elementary School, Gangneung Wonju National University, GIST, Hanguk University of Foreign Studies, Iksan Fire Station, KMA (Anmyeon, Gosan Station), KNJ Engineering, Kyungpook National University, Pusan National University, Seoul National University, Songchon Elementary School, UNIST Satellite Remote Sensing Korea Meteorological Administration (MI), Korea Ocean Satellite Center(KOSC), JMA, NICT NASA Goddard Space Flight Center Brent Holben Tom Eck (AERONET), Jay R. Herman, Abuhassan Nader (Pandora) Scott J. Janz, Matthew G. Kowalewski (GeoTASO), Gordon Labow (MFRSR, UV-MFRSR) NASA Langley Research Center James Crawford, Jay Al-Saadi (GeoTASO, MOS), John Hair (DIAL/HSRL) NASA Ames Research Center Jens Redemann (4STAR) Yonsei University Ja-ho Koo, Mijin Kim, Myungje Choi, Woogyung Kim, Sujung Go, Hana Lee, Heesung Chong, Seoyoung Lee, Hyunkwang Lim 2

3 Development of Satellite RS Capability for AQ Temporal Resolution week day hr min GOCI-2 MODIS GOCI H-8 AMI MISR TEMPO MI TROP OMI OMI GOSAT CAI S-4 GEMS OMPS SCIAMACHY GOME-2 GOME Aerosol Trace Gases Spatial Resolution (km)

4 KORUS-AQ combined assets from the Korean and U.S. atmospheric science communities and their supporting organizations (NIER, NASA, Universities, etc.) to implement an integrated observing system for improving our understanding of Air Quality NASA DC-8 (HSRL, 4 STAR..) LaRC King Air (GEOTASO, MOS) Hanseo King Air [Courtesy of James Crawford] KORUS-OC (1 May 14 June 2016) Model evaluation and improvement, chemical process understanding, Satellite Cal/Val and observing strategies GOCI, MI, Himawari-8, MODIS, OMI, MOPITT.. Broad spatial coverage for key atmospheric components (aerosols, ozone, precursors) Operational Air Quality Forecasts, Regional and Global models of atmospheric composition Air Quality Network, Research Sites, Research Vessels including in situ and remote sensing observations (Aeronet, Pandora, Lidar)

5 Geostationary satellites aerosol observation over East Asia GOCI MI AHI Temporal Resolution GOCI/COMS (KOSC/KIOST, Korea) 1-hour interval for East Asia (total 8 times in daytime) MI/COMS (NMSC/KMA, Korea) 15-min interval for Asia 3-hour interval for FD Channels 8 bands in VIS-NIR (0.5 km) 1 bands in VIS (1 km) 4 bands in IR (4 km) AHI/Himawari-8 (JMA, Japan) 10-min interval for Full Disk 4 bands in VIS-NIR (0.5/1.0 km) 12 bands in IR (2 km) Products AOD, FMF, AE (6 km) AOD (4km) AOD, FMF, AE (6 km) Reference M. Choi et al.(amt 2016) M. Kim et al. (RSE 2014; ACP 2016) H. Lim et al. (KJRS 2016) * Datasets readily available for past years for GOCI and MI. AHI dataset is under processing. 5

6 Case of 25 May 2016 HSRL Height 4STAR AOD (Zenith) GOCI AOD (1 hr) MI AOD (15 min) AHI AOD (10 min) 1~2 km (Median of aerosol extinction profile) All campaign data avg. Mean of case date Range of 16/50/84 percentile HSRL Aerosol type Dusty Mix Smoke Anmyon (36.54N,126.33E) NOAA HYSPLIT Backward trajectory modeling 6

7 Case of 05 June 2016 HSRL Height 4STAR AOD (Zenith) GOCI AOD (1 hr) MI AOD (15 min) AHI AOD (10 min) 1~3 km (Median of aerosol extinction profile) All campaign data avg. Mean of case date Range of 16/50/84 percentile HSRL Aerosol type 7 Fresh Smoke and Smoke

8 Land surface reflectance climatology using a minimum reflectivity technique with multi-year samples Version 1 (Choi et al., AMT, 2016) Version 2 Composite each year/month/hour samples within 6 km 6 km darkest 1-3% at 412 nm Less influence of degradation/calibration issue (pros) Near-real-time retrieval impossible (cons) Composite 5-year/month/hour samples within 500 m 500 m as higher resolution (pros) darkest 1-3% at each channel possibility finding clear pixels increases (pros) Much influence of degradation/calibration issue Hard to reflect surface changes (cons) Near-real-time retrieval possible (pros) V1 Surface reflectance (15 May, 04:30utc, Ch3) V2 Surface reflectance database (15 May, 04:30utc, Ch3) 8

9 Land/Ocean AOD (Total 27/17 AERONET sites, 5yr) GOCI YAER V1 (all QA) GOCI YAER V1 (QA3) GOCI YAER V2 MODIS DT MODIS DB Land Ocean Land AOD N R Median Bias Ratio within EE DT RMSE V1 AllQA V1 QA V DT DB Most statistics show land/ocean algorithm improvement from V1 to V2 EE DT = ±( AOD A ) 9 Ocean AOD N R Median Bias Ratio within EE DT RMSE V1 AllQA V1 QA V DT

10 AOD error analysis from long-term validation No. of each bin samples = 1000, Symbols: 50 percentile, Lines: percentile Reduced bias, increased error in high AOD shorter atmospheric path length, and higher signal of surface in high scattering angle positive bias in NDVI of (urban over China, Osaka, especially) Positive bias and wider error range in low AE (large particle), (could be aerosol model error) positive AOD bias in higher NDVI (open ocean) due to not considering water-leaving radiance 10

11 AOD error analysis from long-term validation Positive bias in high cloud fraction Positive bias in high cloud fraction Positive bias in inhomogeneous AOD 11

12 Spatial distribution: GOCI monthly AOD 12 North China Plane (including Beijing, Tianjin, and Hebei) - Expanded to Korea and Japan in March to May, highest in June-July-August - minimum mean AOD is above 0.4 Korean Peninsula: increased in March and April, highest in June Spatial distribution is well matched with MODIS results (Kim et al. 2007; Levy et al. 2013; Hsu et al. 2013) and VIIRS results (Liu et al. 2014) High AOD over turbid water (a few samples)

13 AHI RGB image AHI Aerosol Products AHI YAER AOD AHI JAXA AOD AHI YAER aerosol type MODIS AOD (Terra,Aqua) VIIRS AOD (suomi NPP) MODIS FMF (Terra,Aqua) AHI YAER FMF AOD range FMF range

14 Monthly validation results AHI YAER vs AERONET Mar, 2016 Apr, 2016 May, 2016 Jun, 2016 Jul, 2016 Analysis period : Spatial colocation : average of AHI pixels within 25km at AERONET sites - (Mar: 29sites, Apr: 42sites, May: 44site, Jun:40 sites, Jul: 26sites) Temporal colocation : average of AERONET data within 5min at satellite measurement time Expected Error (EE) = *AERONET AOD (Levy et al., 2007) 14

15 AHI IR cloud masking in AHI YAER algorithm GOCI YAER V2 GOCI YAER V2 + AHI IR cloud masking Better cloud edge masking Retain high AOD plume AHI IR cloud masking works successfully on GOCI AOD to filter out cirrus or shallow cloud contamination as retaining high AOD well. AHI IR cloud masking results in increased correlation coefficient b/w GOCI and MODIS/VIIRS over ocean (R: with MODIS, with VIIRS) 15

16 Trend Analysis using GOCI & MODIS ( )

17 Application of GOCI YAER AOD to Air quality modeling studies Data assimilation of GOCI AOD with CMAQ Application to the PM 10 (GIST) [Park et al., 2014, ACP] Data assimilation of GOCI & MODIS AOD with WRF-chem Application to the PM 10 (Univ. of Iowa and NCAR) [Saide et al., 2014, GRL] (also carried out during 2016 KORUS-AQ as NRT) [Lee et al., 2016, GMD] Estimation ground-level PM 2.5 from GOCI AOD and GEOS-chem [Xu et al., 2015, ACP] (Dalhousie Univ.) GIST: spatial-temporal kriging to fill the empty data for model time and cloud-masked area 17

18

19 GOCI-2 Spectral bands 250 m for RLA / 1 km for FD Concept of GOCI-2 Aerosol algorithm Courtesy of Dr. Youngje Park (KOSC/KIOST) 19

20 Geostationary Constellation of UV-Vis spectrometer & Meteorological Payload CEOS ACC(Committee on Earth Observaing Satellites Atm. Comp. Constella TEMPO + GOES-R (America) UV-Vis nm TROPOMI, OMPS, VIIRS, APOLLO Himawari 8 FY-2, 3, COMS INSAT GEMS + AMI + GOCI2 GEO KOMPSAT (Asia) UV-Vis nm GMES S4 UVN + FCI + IRS MTG (Europe) UV-Vis-NIR , nm Constellation synergy - Improving spatial and temporal coverage to monitor globalized pollutants & SLCF - Sharing basic requirements on data products and instrument to maintain data quality - Consolidating socio-economic benefit analysis - Supporting QA and CAL/VAL

21 Conclusion GOCI has provided long term dataset on aerosol properties in high spatial and temporal resolution, which has been validated with ground-based AERONET and SONET. Meteorological imager such as MI and AHI Satellite aerosol retrieval algorithm has been under continuous improvement through the evaluation and diagnosis of the AOPs, especially according to aerosol vertical distribution and types. It could provide the key of AOPs-PM relations over wider region. Data assimilation with GOCI aerosol datset improve the accuracy of air quality forecasting. Further application with GOCI AOD was demonstrated to estimate surface PM10 and PM2.5 with correlation coefficients up to 0.8, which then can be used for public health studies. GOCI-2 to be launched in 2019 is expected to provide much more information with its UV channels at 380 nm together with aerosol precursor measurements by GEMS and high temporal resolution observation by AMI. GOCI data is available in near real time(nrt) and the past, 6-yr dataset is readily available upon request. GOCI data are distributed through KIOST website. 21

22 22 Case of 09 June 2016